pois.fe.robust: Robust standard errors of Poisson fixed effects regression

In poisFErobust: Poisson Fixed Effects Robust

Description

Compute standard errors following Wooldridge (1999) for Poisson regression with fixed effects, and a hypothesis test of the conditional mean assumption (3.1).

Usage

pois.fe.robust(outcome, xvars, group.name, data, 
               qcmle.coefs = NULL, allow.set.key = FALSE,
               index.name = NULL)

Arguments

outcome	character string of the name of the dependent variable.
xvars	vector of character strings of the names of the independent variables.
group.name	character string of the name of the grouping variable.
data	data.table which contains the variables named in other arguments. See details for variable type requirements.
qcmle.coefs	an optional numeric vector of coefficients in the same order as xvars. If NULL, coefficients are estimated using glmmML::glmmboot.
allow.set.key	logical. When TRUE (recommended), data will have its key set to group.name, so it may be reordered. This should reduce memory usage.
index.name	DEPRECATED (leave as NULL).

Details

data must be a data.table containing the following:

• a column named by outcome, non-negative integer

• columns named according to each string in xvars, numeric type

• a column named by group.name, factor type

• a column named by index.name, integer sequence increasing by one each observation with no gaps within groups

No observation in data may contain a missing value.

Setting allow.set.key to TRUE is recommended to reduce memory usage; however, it will allow data to be modified (sorted in-place).

pois.fe.robust also returns the p-value of the hypothesis test of the conditional mean assumption (3.1) as described in Wooldridge (1999) section 3.3.

Value

A list containing

• coefficients, a numeric vector of coefficients.

• se.robust, a numeric vector of standard errors.

• p.value, the p-value of a hypothesis test of the conditional mean assumption (3.1).

Author(s)

Evan Wright

References

Wooldridge, Jeffrey M. (1999): "Distribution-free estimation of some nonlinear panel data models," Journal of Econometrics, 90, 77-97.

See Also

glmmboot

Examples

# ex.dt.good satisfies the conditional mean assumption
data("ex.dt.good")
pois.fe.robust(outcome = "y", xvars = c("x1", "x2"), group.name = "id",
               index.name = "day", data = ex.dt.good)
               
# ex.dt.bad violates the conditional mean assumption
data("ex.dt.bad")
pois.fe.robust(outcome = "y", xvars = c("x1", "x2"), group.name = "id",
               index.name = "day", data = ex.dt.bad)

Example output

$coefficients
       x1        x2 
0.9899730 0.9917526 

$se.robust
        x1         x2 
0.03112512 0.02481941 

$p.value
[1] 0.6996001

Warning messages:
1: In pois.fe.robust(outcome = "y", xvars = c("x1", "x2"), group.name = "id",  :
  index.name is deprecated. It is not necessary.
2: In pois.fe.robust(outcome = "y", xvars = c("x1", "x2"), group.name = "id",  :
  allow.set.key is FALSE. TRUE is recommended to avoid copying the data; however, TRUE will sort the data in-place by group.name.
$coefficients
       x1        x2 
0.4800735 2.9866911 

$se.robust
       x1        x2 
0.2864666 1.2743953 

$p.value
[1] 0.02213269

Warning messages:
1: In pois.fe.robust(outcome = "y", xvars = c("x1", "x2"), group.name = "id",  :
  index.name is deprecated. It is not necessary.
2: In pois.fe.robust(outcome = "y", xvars = c("x1", "x2"), group.name = "id",  :
  allow.set.key is FALSE. TRUE is recommended to avoid copying the data; however, TRUE will sort the data in-place by group.name.

poisFErobust documentation built on Feb. 18, 2020, 1:08 a.m.